N 4-7... Long-Run Determinant of the Sovereign CDS spread in emerging countries Abstract Sy Hoa HO* In this paper, we study the long-run determinant of Sovereign CDS spread for eight emerging countries in the 8.Q4-3.Q2 periods. The determinant of sovereign CDS spread is estimated from three macroeconomic factors: the current account, the external debt and the international reserves. Using the Pooled Mean Group cointegration approach, our findings can be summarized as follow: i, the existence of cointegration between these variables indicated above; ii, the coefficients of the current account, the external debt and international reserves are significant in the long-run for all countries; iii, the short-run is significant just for the external debt and the international reserves, not for the current account. Key words: Sovereign Credit Default Swaps, Panel cointegration, Pooled Mean Group. JEL classification: F34, G15, C32, C33 *CEPN (UMR-7234 CNRS), Université Paris Nord, Paris Sorbonne Cité, 99 Avenue Jean Baptiste Clément, 93430 Villetaneuse, France.
Long-Run Determinant of the Sovereign CDS spread in emerging countries HO Sy-Hoa February 28, 4 Abstract In this paper, we study the long-run determinant of Sovereign CDS spread for eight emerging countries in the 8.Q4-3.Q2 periods. The determinant of sovereign CDS spread is estimated from three macroeconomic factors: the current account, the external debt and the international reserves. Using the Pooled Mean Group cointegration approach, our findings can be summarized as follow: i, the existence of cointegration between these variables indicated above; ii, the coefficients of the current account, the external debt and international reserves are significant in the long-run for all countries; iii, the short-run is significant just for the external debt and the international reserves, not for the current account. Key words: Sovereign Credit Default Swaps, Panel cointegration, Pooled Mean Group. JEL classification: F34, G15, C32, C33 CEPN (CNRS-UMR 7234), University of Paris North, Paris Sorbonne Cité, France 1
1 Introduction and Literature Credit Default Swap (CDS) is an insurance contract between a seller and a buyer that allows investors to buy protection against default. The buyer pays a fee to the seller (called CDS premium or spread), in exchange, the buyer of the CDS receives compensation from the seller in the event of default. The aim of sovereign CDS is to insure the sovereign bonds buyer against the default of government bond. The larger the CDS spread on a government bond, the higher the risk of default on that bond by the issuing government. The short-selling of CDS can lead the sovereign bond downgrade, which led to disorder of the CDS market and the global financial crisis in 7. Additionally, Chan-Lau (3, 6) demonstrates the relationship between the CDS spread and sovereign default probability. The default probability can be computed and predicted from the CDS spread. Therefore, the CDS spread has been a proxy of country risk since then. We start by reviewing the various studies of the determination of sovereign risk (e.g the sovereign CDS or default probability of sovereign) by fundamental macroeconomics variables. Bulow and Rogoff (1988) show that exchange rate variations have a direct impact on a country s terms of trade, which may affect the ability of the country to generate dollar revenue and make payments on its external debt. Hernandez-Trillo (1995) indicates that the default probability of sovereign depends on: the degree of openness, a degree of "unluckiness", international reserves and the risk-free interest rate. According to his model, he creates a spread index over LIBOR and debt service ratio to determine the sovereign default probability, and defines the default probability as a function of the cost of default. His results analyse for 33 debtor emerging countries in the period 1970-1988 that liberalization policies decreases the probability of default by both raising the GDP and increasing openness. A country default will induce a loss in access to future credit. The unluckiness is also important when explaining the probability of default. This model displays a negative effect of openness and international reserves on the default probability. Cantor and Packer (1996) regress a country s ability and willingness-to-service its debt for a panel of developing countries. They find six variables affecting the sovereign credit rating which are per capita income, GDP growth, inflation rate, external debt, default history and an economic development indicator. Additionally, Mellios and Paget-Blanc (6) determine two other factors that are government income and change in the real exchange rate. These variables have a positive impact on default probability except for the inflation rate. Baek et al. (5) show the link between sovereign risk in emerging countries and macroeconomic variables such as government budget balance and current account balance. Georgievska et al. (8)use the Bayesian approach to study a panel of 124 emerging countries during the 1981-2 period. Their results report three classified variables explaining sovereign default: total debt to GDP ratio and Export to GDP ratio represent solvency variables, international reserves to GDP ratio express liquidity and currency account balance to GDP ratio and imports to GDP ratio variables represent macroeconomic variables. Ramos-Francia and Rangel (2) create a sovereign spread index as the difference between the long-term government bonds yield and the 10-year US Treasuries yield. They test the relationship of this index with macroeconomic variables such as inflation, economic growth, fiscal and current account deficits, international reserves and nominal exchange rate variations. These results illustrate that the international reserves and exchange rate appreciations are associated with lower default risk in emerging markets. IMF-Report (3) introduces the determinant of the CDS spread by regressing it on various economics and financial explanatory variables. There are two variables that would be expected to increase the spread; debt-to-gdp ratio and GDP growth 2
rate whereas international reserves would reduce it. Additionally, this report suggests global or region-specific explanatory variables such as VIX (S&P 500 index), global equity return (1-month US Treasury) and Funding cost (3-month LIBOR-OIS). Delatte et al. (2) use a vector error correction model (VECM) in order to test the long-run cointegration between the CDS premia and the bond spread for the European countries in the 8.M9-0.M7 period. The limitations of previous papers have not examined this subject for emerging countries in the long-run and short-run. Figure 1, the sovereign CDS 5-year spread of eight emerging countries is fairly homogeneous in 8.Q4-3.Q2 periods 1. In other word, the country risk for all countries of this panel is homogeneous in this period. Furthermore, the sovereign CDS 5-year spread is the most traded in the CDS market. That is why we use this proxy in the paper to determine long-run sovereign CDS spread. In this paper, we contribute the empirical results of the long-run and the short-run relationship between the sovereign CDS spread and fundamental macroeconomics for this panel group. We propose the explanatory variables in this paper are the current account to GDP ratio, the external debt to GDP ratio and the international reserves to GDP ratio such as in the article of Georgievska et al. (8) and IMF-Report (3). Figure 1: Sovereign CDS 5-year spread 700 600 500 400 300 100 0 08.q4 09.q1 09.q2 09.q3 09.q4 10.q1 10.q2 10.q3 10.q4 11.q1 11.q2 11.q3 11.q4 12.q1 12.q2 12.q3 12.q4 13.q1 13.q2 Bra_CDS5 Mala_CDS5 S Kor_CDS5 Thai_CDS5 Tur_CDS5 S Afri_CDS5 Indo_CDS5 Mexi_CDS5 Source: Reuters The remainder of this paper is organized as follows: in the next section we present the econometric methodology to exam the long-run, short-run of the determinant of sovereign CDS spread and empirical results. The last section is a conclusion. 2 Econometric approach and Empirical results Data Description In this paper, we use quarterly data from 8.Q4 to 3.Q2 for eight emerging countries: Brazil, Malaysia, South Korean, Thailand, Turkey, South Africa, Indonesia and Mexico. Sovereign 1 We focus on the 8q3-3q4 periods for reason of data availability. 3
CDS 5-year spread in log denoted LCDS is taken from Reuters covered from daily data. CA is the ratio of the current account to GDP, ED is the ratio of the external debt to GDP. The two are found in Central Bank of each country. RES is the international reserve to GDP ratio (IMF International Financial Statistic). In order to test the long-run and short-run relationship between the sovereign CDS spread and the explanatory variables for the emerging panel, we apply the panel unit root and the panel cointegration test. The panel unit root aims to verify that all variables are integrated with the same order. The panel cointegration test is to study the existence of cointegration between all variables. These tests are the necessary condition to test the panel long-run estimation. 2.1 Panel unit root Before testing cointegration, we verify that all variables are integrated with the same order. We employ the first generation panel unit root tests of Levin et al. (2), Im et al. (3) which are denoted LLC and IPS and the second generation one of Pesaran (7) which is denoted CIPS. The null hypothesis is that the series contains a unit root. Levin et al. (2) assumes that all panels have the same autoregressive parameter, but Im et al. (3) allows each panel to have its own autoregressive parameter. The panel unit root of Pesaran (7) takes into account crosssectional dependence. The panel unit root test of all variables and the first difference its report in Figure 2. Figure 2: Panel unit root for sovereign CDS, current account, external debt and exchange rate Series Levin, Lin, Chu Im, Pesaran, Shin Pesaran (CIPS) LCDS -9.2466***(0.0000) -3.4171***(0.0003) -1.671 (0.564) CA -0.9600(0.1685) -1.3269*(0.0923) -2.008(0.222) RES -0.2669(0.3948) -0.7367(0.2306) -1.888(0.332) ED -1.3043*(0.0961) 0.0450(0.5179) -1.918(0.303) LCDS -5.1157***(0.0000) -4.7774***(0.0000) -2.559** (0.011) CA -6.8535***(0.0000) -6.5774***(0.0000) -2.944*** (0.003) RES -4.3401***(0.0000) -6.4481***(0.0000) -2.817***(0.001) ED -0.5025(0.3076) -6.6663***(0.0000) -3.339*** (0.000) Notes: is first difference, ***, **, * means that we reject the null hypothesis of unit root at the 1%, 5%, 10% level. P-value in parentheses For the first generation of panel unit root: The variables LCDS is stationary, do not include a unit root with the LLC, IPS test at the 1% level. Concerning the variables CA, this variable is stationary with IPS test at the 10% level, but not with LLC test. The variable RES is not stationary with the both LLC, IPS test. The variable ED is stationary with LLC test at the 10% level, but not IPS test. The first difference of all series is stationary for LLC, IPS, CIPS test. 4
The second generation CIPS test solves the biased results in the first generation. Focusing on the CIPS test, all variables in level are integrated with the same order I(1). Therefore, we can test the existence of cointegration relationship among LCDS, CA, RES, ED. 2.2 Panel cointegration Test The test of panel cointegration proposed by Pedroni (1999) and Kao (1999). Pedroni (1999) presented seven statistics: Panel-v, Panel-ρ, Panel-PP, Panel-ADF (the four statistics are based on within-dimension), and Group ρ-statistic, Group PP-statistic, Group ADF-statistic (the three statistics are based on between-dimension). Kao (1999) tested the residual estimation based on Augmented Dickey-Fuller test. The null hypothesis of the both tests is no cointegration. The results are shown in Figure 3 and Figure 4 in below: Figure 3: Panel cointegration test: Pedroni test Pedroni(1999) Statistic Prob. Panel-v 0.7316 0.4977 Panel-ρ -1.3928 0.1267 Panel-PP -2.300*** 0.0030 Panel-ADF -1.5619*** 0.0026 Group ρ-statistic 0.1727 0.5686 Group PP-statistic -2.9813*** 0.0014 Group ADF-statistic -3.1012*** 0.0010 Figure 4: Panel cointegration test: Kao test Kao(1999) Statistic Prob. ADF 1.4535* 0.0730 ***, **, * means that we reject the null hypothesis of non-cointegration at the 1%, 5%, 10% level Figure 3: Regarding the Panel-PP, Panel-ADF (within-dimension) statistic and Group PPstatistic and Group ADF-statistic (between-dimension), the results show that we have 4/7 tests that reject the null hypothesis of no cointegration at the 1 % level, except Panel-v, Panel-ρ and Group-ρ which are not significant. More that, the results in Figure 4 confirm that we reject the null hypothesis of no cointegration with ADF statistic at the 10 % level. Therefore, we can conclude that there exists cointegration among the sovereign CDS and the current account, the external debt and the international reserves. 5
2.3 Panel cointegration Estimation : Pooled Mean Group estimation In order to test Panel cointegration estimation, we use Pooled Mean Group (PMG) estimation approach. Pooled Mean Group estimates the presence of dynamic long-run, short-run based on the methodology of Pesaran et al. (1997, 1999). We analyse the role of the current account, the external debt and the international reserves with sovereign CDS spread for all countries in this panel. The techniques of PMG are used in the recent article of Martinez-Zarzoso and Bengochea- Morancho (4); Bangake and Eggoh (2). The approach we adopted here is the determination of long-term relationship among LCDS, CA, RES, ED, LER. Here, we investigate cointegration estimation using an empirical formalization using panel data. We assume that the long-run sovereign CDS function is given by: LCDS it = α 0t + α 1t CA it + α 2t RES it + α 3t ED it + µ i + ɛ it (1) Where LCDS is sovereign CDS spread in log; CA is ratio of the current account to GDP; RES is ratio of the international reserves to GDP; ED is ratio of the external debt to GDP; µ i is the group-specific effect; i = 1, 2,..., 8 is the number of countries; t=1,2...t is number of periods. We assume that all these variables are I(1) and co-integrated for individual countries, leading the error term an I(0) process for all i. The ARDL(1,1,1,1) dynamic panel specification of equation (1) is LCDS it = δ 10i CA it + δ 11i CA i,t 1 + δ 20i RES it + δ 21i RES i,t 1 + δ 30i ED it + δ 31i ED i,t 1 + λ i LCDS i,t 1 µ i + ɛ it (2) The error correction (ec) of equation (2) is : LCDS it = φ i (LCDS i,t 1 α 0i α 1i CA it α 2i RES it α 3i ED it ) + δ 11i CA it + δ 21i RES it + δ 31i ED it + ɛ it (3) Where φ i = (1 λ i ), α 0i = µ i, α 1i = δ 10i + δ 11i, α 2i = δ 20i + δ 21i, α 3i = δ 30i + δ 31i, 1 λ i 1 λ i 1 λ i 1 λ i The error-correction of adjustment parameter is φ i, the long-run coefficient is α 1i, α 2i, α 3i for CA, RES, ED respectively. In long-run equilibrium, the coefficient of error-correction φ i must be negative and significant (Pesaran et al., 1997, 1999). PMG allows for heterogeneous common long-run for all countries, and short-run dynamic the sovereign CDS, the current account to GDP, the external debt to GDP and the international reserves to GDP. The results of the long-run and short-run of the current account to GDP, the external to GDP, the international reserves to GDP and the error correction coefficient report in Figure 5. 6
Figure 5: Panel cointegraion estimation: Pooled Mean Group Variable Long-run coef. Variable Short-run coef CA -3.535*** CA 3.148** (1.139) (1.237) ED 2.470*** ED 1.779** (0.319) (0.817) RES -5.610*** RES -1.803* (0.873) (1.923) constant 2.611*** (0.560) ec -0.437*** (0.112) Notes: Standard errors in parentheses ***, **, * means that the coefficient is significant at the 1%, 5%, 10%, level In long-run equilibrium, the estimated coefficient of error-correction is negative -0.437 and significant at the 1% level as expected. The estimation results show: the coefficient of the current account to GDP ratio is negative -3.535 and significant at the 1 % level for all countries. This result reflects that when all countries improve their current account, the sovereign CDS spread will decrease because the government has more money to repay its debt to lenders. The coefficient of the external debt to GDP is positive, 2.470 and significant for all countries at the 1 % level as expected. This finding shows that evidently if the external debt increases, then sovereign default probability will increase. That affects badly the sovereign CDS spread because the sovereign CDS spread is a country risk indicator. When the economic situation is not good, or the external debt has risen, then sovereign CDS spread will rise. The coefficient of international reserve is negative, -5.610 and significant at the 1 % level as expected. This coefficient implies that if the country faces "bad" economic situation, in order to prevent default, the government can use international reserves to repay its debt. Thus, if the government has more reserves, then the sovereign default probability will decrease. We could also go back to Figure 1, the homogeneous of the sovereign CDS spread of eight countries, and notice that we have a significant long-run effect of three important variables affecting the sovereign CDS spread, which are representative for the government s solvency, liquidity and macroeconomics variables(georgievska et al., 8). Our results suggest the long-run coefficient of the international reserve is the highest relatively to the two other coefficients. From the policy s point of view, in order to reduce the country risk in the long-run, governments should focus more on increasing their reserves than on solving two factors: the external debt and the current account. The estimation of the short-run coefficients are mentioned in Figure 5 for the global panel and in Figure 6 for each country. 7
Figure 6: Panel cointegration estimation: short-run for each country in PMG Variable Bra Mala S-Kor Thai Tur S-Afri Indo Mexi ec -0.315** -0.150* -0.813*** -0.120-1.000*** -0.231*** -0.362** -0.505** (0.140) (0.0792) (0.205) (0.0847) (0.169) (0.0808) (0.149) (0.) CA -0.409-0.902 5.122** 0.479 8.352*** 0.989 5.714 5.838 (6.739) (1.716) (2.383) (1.013) (1.257) (1.349) (15.89) (5.564) RES -2.053-0.791 2.529-0.486-3.918*** 3.867* -13.83 0.263 (3.219) (0.599) (2.679) (0.760) (1.269) (2.161) (15.94) (2.819) ED 3.933 2.218*** -0.739 3.982 0.708-1.634** 4.634 1.132 (3.076) (0.651) (1.507) (5.448) (0.439) (0.726) (5.338) (1.818) Constant 2.132** 1.760* 6.033*** 1.770 3.601*** 1.036*** 1.767** 2.787** (0.939) (0.925) (1.625) (1.298) (0.712) (0.386) (0.789) (1.123) Observations 144 144 144 144 144 144 144 144 Standard errors in parentheses ***, **, * means that the coefficient is significant at the 1%, 5%, 10%, level For the all countries, three short-run coefficients are significant statistically. But in as expected, the external debt to GDP and the international reserves to GDP are significant as theoretical economics at the 5% and 1% level respectively. As Figure 6 show, the short-run coefficient of the international reserves is -3.918 significant as expected at the 1% level for Turkey. Likewise, the short-run coefficient of external debt is 2.218 significant at the 1% level for Malaysia as expected. The coefficient of the current account to GDP is not significant vis-à-vis theoretical economics for the global panel and each country also. 3 Conclusion This paper provides the link between the sovereign CDS spread and the current account, the external debt and the international reserves for eight emerging countries by using the new technique of the Pooled Mean Group. Our results suggest a great long-run effect of the current account, the external debt and the international reserves on the sovereign CDS spread of the whole panel. The sign of explanatory variables is confirmed with the previous studies of Georgievska et al. (8) and IMF-Report (3). Our results also indicate that in the short-run, the external debt and the international reserves are significant for eight emerging countries. Especially, the shortrun coefficient of the international reserves, the external debt is significant for Turkey, Malaysia respectively. The current account is not significant in the short-run for all countries. Acknowledgement I would like to thank Jean Michel COURTAULT, Natalie REY, Idir HAFRAD, Jamel SAADAOUI, Serge HERBILLON for their discussions. Financial support from CEPN, Ecole Erasme is acknowledged. 8
References Baek, I.-M., A. Bandopadhyaya, and C. Du (5). Determinants of market-assessed sovereign risk: Economic fundamentals or market risk appetite? Journal of International Money and Finance 24 (4), 533 548. Bangake, C. and J. Eggoh (2). Pooled mean group estimation on international capital mobility in african countries. Research in Economics 66 (1), 7 17. Bulow, J. I. and K. Rogoff (1988). Sovereign debt: Is to forgive to forget? American Economic Review 79, 43 50. Cantor, R. and F. Packer (1996). Determinants and impact of sovereign credit ratings. Economic Policy Review (Oct), 37 53. Chan-Lau, J. (3). Anticipating credit events using credit default swaps, with an application to sovereign debt crises. IMF Working Papers 03/106, International Monetary Fund. Chan-Lau, J. (6). Market-based estimation of default probabilities and its application to financial market surveillance. IMF Working Papers 06/104, International Monetary Fund. Delatte, A.-L., M. Gex, and A. López-Villavicencio (2). Has the cds market influenced the borrowing cost of european countries during the sovereign crisis? Journal of International Money and Finance 31 (3), 481 497. Georgievska, A., L. Georgievska, A. Stojanovic, and N. Todorovic (8). Sovereign rescheduling probabilities in emerging markets: a comparison with credit rating agencies ratings. Journal of Applied Statistics 35 (9), 1031 1051. Hernandez-Trillo, F. (1995). A model-based estimation of the probability of default in sovereign credit markets. Journal of Development Economics 46 (1), 163 179. Im, K. S., M. H. Pesaran, and Y. Shin (3). Testing for unit roots in heterogeneous panels. Journal of Econometrics 115 (1), 53 74. IMF-Report (3). Chapter 2: A new look at the role of sovereign credit default swaps. Global Financial Stability Report: Old Risks, New Challenges. Global Financial Stability Report. International Monetary Fund. Kao, C. (1999). Spurious regression and residual-based tests for cointegration in panel data. Journal of Econometrics 90 (1), 1 44. Levin, A., C.-F. Lin, and C.-S. James Chu (2). Unit root tests in panel data: asymptotic and finite-sample properties. Journal of Econometrics 108 (1), 1 24. Martinez-Zarzoso, I. and A. Bengochea-Morancho (4, January). Pooled mean group estimation of an environmental kuznets curve for co2. Economics Letters 82 (1), 121 126. Mellios, C. and E. Paget-Blanc (6). Which factors determine sovereign credit ratings? European Journal of Finance 12 (4), 361 377. 9
Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics 61 (0), 653 70. Pesaran, M. H. (7). A simple panel unit root test in the presence of cross-section dependence. Journal of Applied Econometrics 22 (2), 265 312. Pesaran, M. H., Y. Shin, and R. J. Smith (1997). Estimating long-run relationships in dynamic heterogeneous panels. DAE Working Papers Amalgamated Series 9721. Pesaran, M. H., Y. Shin, and R. J. Smith (1999). Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American Statistical Association 94, 621634. Ramos-Francia, M. and J. G. Rangel (2). Revisiting the effects of country specific fundamentals on sovereign default risk. Economics Bulletin 32 (4), 3008 3016. 10